Virtualization Services: Accelerated I/O Support in Multi-Core Systems
نویسندگان
چکیده
Virtualization services permit I/O subsystems and peripheral devices to be virtualized by placing select functionality on specialized cores and/or on cores situated ‘closer’ to devices. The approach is used to implement self-virtualized I/O (SVIO), which off-loads certain virtualization functionality onto the cores available on the I/O device, accelerating I/O functions, efficiently using key performance-limiting resources in multicore systems, i.e., memory and I/O bandwidth, and exploiting the parallelism inherent in multi-core architectures. This paper evaluates a concrete instance of self-virtualized I/O, a selfvirtualized network interface (SV-NIC), targeting the high end NICs used in datacenters. Experimental evaluations of the SVNIC in a prototyping environment using an IXP2400-based ethernet board show high scalability in terms of the numbers of virtual interfaces (VIFs) offered to guests, and up to ∼77% improvements in throughput and ∼53% reductions in latency, compared to the current standard virtualized device implementations on hypervisor-based platforms. Beyond such performance advantages, the generality of virtualization services and their use for implementing enhancements to standard services are demonstrated with a storage service that provides location transparent access to block devices for guest VMs, and with enhancements to a network service that provide per-VM, prioritybased servicing of virtual network devices.
منابع مشابه
Improving I/O Resource Sharing of Linux Cgroup for NVMe SSDs on Multi-core Systems
In container-based virtualization where multiple isolated containers share I/O resources on top of a single operating system, efficient and proportional I/O resource sharing is an important system requirement. Motivated by a lack of adequate support for I/O resource sharing in Linux Cgroup for high-performance NVMe SSDs, we developed a new weight-based dynamic throttling technique which can pro...
متن کاملEfficient and Scalable Paravirtual I/O System
The most popular I/O virtualization method today is paravirtual I/O. Its popularity stems from its reasonable performance levels while allowing the host to interpose, i.e., inspect or control, the guest’s I/O activity. We show that paravirtual I/O performance still significantly lags behind that of state-of-the-art non-interposing I/O virtualization, SRIOV. Moreover, we show that in the existin...
متن کاملSplitX: Split Guest/Hypervisor Execution on Multi-Core
Current virtualization solutions often bear an unacceptable performance cost, limiting their use in many situations, and in particular when running I/O intensive workloads. We argue that this overhead is inherent in Popek and Goldberg’s trap-and-emulate model for machine virtualization, and propose an alternative virtualization model for multi-core systems, where unmodified guests and hyperviso...
متن کاملPolicy Model for Sharing Network Slices in 5G Core Network
As mobile data traffic increases, and the number of services provided by the mobile network increases, service load flows as well, which requires changing in the principles, models, and strategies for media transmission streams serving to guarantee the given nature of giving a wide scope of services in Flexible and cost-effective. Right now, the fundamental question remains what number of netwo...
متن کاملTransmission Layer ( Byte Addressable )
Virtualization-based cloud computing has dominated today’s data centers by supporting consolidated servers, converged infrastructures, horizontal scalability, and cost of ownership reduction. As the demands of cloud-based services continue to grow, numerous enabling technologies have been developed such as system virtualization (e.g., VM, container and NFV), software-defined management (for com...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2010